CODEWITHSUNDEEP
c++opencv
Newb
Newb

Reputation: 111

Convert data from opencv effectively

I'm writing a program use both OpenCV and Kinect v2 SDK, where I met an awkward problem. I obtained vectors of Point2i（class consist of int X,Y) as indices of pixels, and I need to convert it into depthSpacePoint（a struct consisting of float32 X，Y）, which is simply depthSpacePoints.X = Point2i.X; or the other way around.

There are so many elements, writing a for loop would be too slow.

So I wonder if there any method like memcpy or other to do it faster?

Here is a brief of my code:

cv::findnonZeros(Mat A, vector<cv::Point2i> B);

vector<depthSpacePoint> C(B.size());

for(int i= 0; i < B.size(); ++i){
C[i].X = b[i].x;
C[i].Y = B[i].y;
}

icoordinateMapper(C.size(),C,...,...);

<- this is the reason why I convert it, because this library function only takes its own data type depthSpacePoint;

Upvotes: 2

Views: 156

Answers (1)

Miki
Miki

Reputation: 41775

You can implement your own version of cv::findNonZero that directly creates a std::vector<depthSpacePoint>.

It's just a matter of scanning (efficiently) the matrix and checking a condition.

void findNonZeroSpacePoints( InputArray _src, std::vector<depthSpacePoint>& pts )
{
    // Get the src matrix, be sure it's of correct type
    Mat src = _src.getMat();
    CV_Assert( src.type() == CV_8UC1 );        

    // If src is all zero, return
    int n = countNonZero(src);
    if(n == 0) {
        pts.clear();
        return;
    }

    // Allocate pts
    pts.resize(n);        

    // Efficiently scan matrix and append points
    depthSpacePoint* ppts = pts.data();
    for( int i = 0; i < src.rows; i++ )
    {
        const uchar* bin_ptr = src.ptr(i);
        for( int j = 0; j < src.cols; j++ )
            if( bin_ptr[j] )
                *ppts ++ = depthSpacePoint(float(j), float(i));
    }
}

Upvotes: 1

Related Questions